Motion Features

From FourMs

Jump to: navigation, search

Contents

The purpose of this document

This document is made as an aid to structure the ways we work with motion data. We see the need to agree on some key features in musical movement, similarly to some of the musical features that are commonly extracted from audio or midi data in the music information retrieval community. This list is continuously edited and updated. Some of the main questions here are:

  • What features can be extracted from musical movement?
  • What features are relevant in various musical contexts?
  • What features can be extracted from different data sources?
  • How do we consider the temporal and spacial distribution simultaneously?
  • How can these low-level features be related to perceptual features?
  • What are the perceptual features of movement?


Quantitative Motion Features

Representation

  • Marker positions
  • Limb/bone positions
  • Joint positions
  • Joint angles
  • Kinematic chains

Posture Features

  • Body extension (the distance between extremeties)
  • Balance (Extremeties pointed to either side of the body, or equally distributed)


Segment Features

By looking at the movement of a single point in time, the following features could be of interest (initially defined for Wacom-data, need to be revised for mocap)

  • Duration of the segment
  • Sub-segment count (How would one segment the segments?)
  • Length of the movement in the segment
  • Average velocity
  • angle sum
  • curviness : angle sum / duration
  • growth
  • pressure-sum
  • intensity
  • air-distance
  • detour

NB! These are preliminary descriptors, and have not yet been compared to other movement descriptors


How should we extract the segment from the stream of motion data?

  • By looking at the points in time where the acceleration is low?
  • A running time-window?

Temporal/Frame-by-frame features

Single marker

  • Velocity
  • Acceleration
  • Jerkiness

Multiple markers

  • Total velocity
  • Total acceleration
  • Total jerkiness
  • Energy


Qualitative/Perceptual Motion Features

Just a first attempt:

  • Energy
  • Smoothness
  • Determination
Personal tools